5
\$\begingroup\$

Considering an NPN BJT transistor, I have read in some texts before that for a more stable control of collector current Ic by the base-emitter voltage Vbe, the effect of Vce on the width of base region is minimized. So Ic Vce curves are better to be flatter.

But how is that achieved in construction? I've read once that, it is done by making the emitter side highest doping concentration, then less concentration the base side and even less doping concentration at the collector side. Is that correct? And how does this minimize the Early effect?

\$\endgroup\$
  • \$\begingroup\$ This might be correct, or not, just putting it out here. My teacher taught me that the Early effect exists because of a small parasitic series resistance at the collector. If what he said is true, then in order to make the Early effect smaller, you should decrease the parasitic resistance. \$\endgroup\$ – Harry Svensson Dec 22 '17 at 21:05
  • \$\begingroup\$ @HarrySvensson the Early effect exists because of a small parasitic series resistance at the collector. Hmm, maybe that teacher was confused with the small signal representation of the Early effect which is a resistor in parallel with the gm*Vbe current source in the Hybrid Pi Model: en.wikipedia.org/wiki/Hybrid-pi_model Physically the Early effect is modulation of the base width and that causes Ic to increase as Vce increases. \$\endgroup\$ – Bimpelrekkie Dec 22 '17 at 21:33
7
\$\begingroup\$

What is the Early effect?

Indeed it has to do with influence of Vce on the Base region. A higher Vce increases the size of the Base-Collector depletion region. This depletion region partly expands into the Base region making it smaller.

See this illustration, (a) shows the transistor in forward mode with a certain Vce. (b) shows the same but with a higher Vce.

enter image description here

This results in a smaller Base region and an increase in current.

What can we do against this?

We cannot prevent the BC depletion region from expanding but we can make the effect smaller by increasing the doping of the Base and Collector regions. A property of a PN junction is that the depletion region decreases in size as the doping levels increase. So indeed, higher doping levels can help however that has other implications.

The current amplification beta of a BJT is determined by the ratio of the dopings. If the doping levels of Base and Collector are increase, the Emitter doping must be increased by the same factor to maintain the same beta. Also the Emitter must have the highest doping level. If the Collector had a higher doping level than the Emitter, it would become the Emitter! and the Emitter would be the collector (or the "reverse beta" would become higher than the "forward beta" and that would be silly).

Increasing doping levels has another disadvantage, as the depletion regions become smaller, the maximum Vce before breakdown decreases.

You could make the Base region larger but that will decrease beta as the minority carriers in the base will have more chance to recombine. A high beta BJT relies on having a short Base region.

So it is a compromise, high-beta transistors suffer more form the Early effect than low beta transistors. So you could also compromise on beta for less Early effect.

What else can we do?

As an IC designer I cannot do all the things I mention above, the fabrication process for the NPNs I have available is fixed. The NPNs are what they are and I have to deal with it. So I have to use circuit solutions.

Like cascoding:

schematic

simulate this circuit – Schematic created using CircuitLab

Here Qcasc takes care of most of the Vce so it will suffer from the Early effect. However, Qcasc does not set Ic, Ic is set by Q1 and Q1 has a nice fixed Vce of around 0.7 V.

A disadvantage of cascoding is that the minimum Vce will be larger than using only Q1 of course so cascoding is not always an option.

An advantage of cascoding is that the Miller effect become much less dominant as the voltage gain from the base to the collector of Q1 is only about 1 (one). For high frequency (RF) amplifiers this advantage might be the only reason why cascoding is done. Then the Early effect isn't really a problem but the bandwith of Q1 is. Cascoding is needed to get and use the full available bandwith.

\$\endgroup\$
  • \$\begingroup\$ You wrote: "A property of a PN junction is that the depletion region decreases in size as the doping levels increase." This is the part Im trying to understand. Does that mean EB junction is made wider than the CB junction? Does it affect the base width? Most importantly I dont get the logical connection between "doping levels/the depletion/base widths" and the Early effect. Is thinner base region in construction means less Early effect? Or the other way around? I understand what Early effect is about but Im wondering the construction effects such as doping variations and initial base width. \$\endgroup\$ – user16307 Dec 22 '17 at 21:34
  • \$\begingroup\$ More doping means smaller depl. region, that's because a certain amount of charge (due to the doping) has to be in that depl. region to "counteract" the reverse voltage across the PN junction in reverse. Note that the BE junction is in forward mode! It will always be smaller as the doping of E and B is always higher than the doping of C. The BE junction is not that relevant for Early effect as it is in forward mode. It does not change when Vce changes. \$\endgroup\$ – Bimpelrekkie Dec 22 '17 at 21:38
  • \$\begingroup\$ Is thinner base region in construction means less Early effect? No, it will be more as a thin base is easier to make even smaller by that CB depl. region pushing its way in. I explained what can be done in my answer. Any change that improves Early effect has a negative consequence (less beta, lower Vce_max). There is no change you can make to only improve Early effect without affecting other parameters. At least, not as far as I know. \$\endgroup\$ – Bimpelrekkie Dec 22 '17 at 21:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.